1 INTRODUCTION

Cardiovascular or cerebrovascular diseases (CVD) remain the leading cause of mortality for people in China and many other countries around the world, including atherosclerotic cardiovascular disease (ASCVD), heart failure (HF), atrial fibrillation (AF), myocardial infarction (MI), and stroke.^{1, 2} According to the survey from 2015 to 2018 by National Health and Nutrition Examination, about 6 million Americans over the age of 20 years had HF with a fairly high number of black males and females.² In China, CVD is increasingly threatening people's health²; this is lightened by the rapid economic development and the change in lifestyle over the past three decades. Besides, CVD mortality is significantly higher in the population of Northern China compared with Southern China.^{3, 4} Although CVD mortality is as serious as the mortality of malignancies, it commonly can be prevented or well controlled by early interventions.⁵ Hence, early warning and identification of subjects at risk is pertinent for preventing and treating of CVD.

Many risks prediction models and equations have been put forward and shown to have high performance in different populations.^6-9 Well-known prediction models for Western populations include the Framingham risk equations for CVD, which is developed based on the populations in America¹⁰ and the Pooled Cohort Equations for estimating the10-year risk of ASCVD for White and African American populations.¹¹ For United Kingdom, the cardiovascular disease risk algorithm (QRISK) has been established as a new 10-year CVD risk score in the derivation cohort.¹² For Asia, Barzi et al.⁹ developed a cardiovascular risk prediction tool. Chien et al.¹³ constructed a 15-year prediction model for predicting incidence of stroke in Chinese adults. The China-PAR project established the Chinese ASCVD risk estimation tool (Prediction for ASCVD Risk in China).¹⁴ However, theses prediction models all derive from cohorts that did not take the physical activity of participants into account.

Physical activity represents a cornerstone in primary level of prevention for CVD, compared with participants with no physical activity, even low intensity of physical activity was correlated with the decreased risk of mortality.² It has been shown to have protective effects in several tissues; in the brain, it has been shown to be positively associated with the total brain volume and cortical thickness in temporal areas such.¹⁵ It has been shown to reduce mortality and morbidity of CVD and importantly, intensity of physical activity appears be beneficial for female life expectancy.¹⁶ Increasing time spent in low-intensity physical activity protects adolescents from effects of long-term sedentary lifestyle.^{17, 18} Intervention measures are needed to promote regular physical activity in school to decrease the sedentary time for child and teenager.¹⁹ Study also shows that moderate physical activity was inversely associated with coronary heart disease.^{20, 21} Various evidence has shown that physically active subjects have a lower incidence of CHD and CVD.²² Thus, physical activity epidemiology has even been born under this circumstance.²² However, in previous risk prediction models, physical activity has not been considered as a factor.^{14, 23} Therefore, it is critical to establish new tools for prediction of CVD that includes physical activity.

The Asymptomatic Polyvascular Abnormalities in Community (APAC) cohort as a population-based study was designed to verify the prevalence and correlations of symptomatic polyvascular abnormalities in the ordinary population.²⁴ The study includes 5440 participants with 40.1% women with age over 40 years. The APAC evaluated the correlation between retinal nerve fiber layer changes, retinal microvascular signs, intracranial arterial stenosis, and eventually study the occurrence and development of a cardiovascular or cerebrovascular event. In the APAC study, assessment for the intensity of physical activity were conducted based on the International Physical Activity Questionnaire (IPQA) short form for all participants^25-27 under the instruction of professional and well-trained researchers. The Kailuan physical activity cohorts used in our study were created from the individuals in APAC study by excluding those who have previously reported HF or AF, and the details can be found in flow chart in Figure 1. Hence, in this study, we first used the participants from the APAC study,²⁴ to develop a new prediction equation that involves physical activity (PA equation) for prediction of CVD events. Second, we validated our PA equation by a 10-fold cross validation and compared it with the China-PAR project by estimating the area under the receiver operating characteristic (ROC) curves (AUC).

2 RESULTS

2.2 Baseline characteristics of participants by physical activity

Baseline characteristics of participants based on physical activity were presented in Table 2, and the p value less than 0.05 represents the significant difference of characteristics among three levels intensity of physical activity groups. As described above, we classified subjects into 3 classes based on the intensity of physical activity. 40%, 25% and 35% subjects had low-intensity activity, moderate-activity and high-intensity activity levels, respectively. Subjects with high-intensity activity were older than moderate and low-intensity activity (p value < 0.001). Low-intensity activity subjects were more likely to be current smoker and drinker (p value < 0.001).

TABLE 2. Baseline characteristics based on intensity of physical activity.

Characteristics	Total (N = 5260)	Low intensity (N = 2104)	Moderate intensity (N = 1339)	High intensity (N = 1817)	p Value
Sex (Men/Women)	3135/2125	1301/803	701/638	1133/684	<0.001a, c
Age, mean (SD), y	55.0$$ \,{\pm}\, $$11.7	52.6$$ \,{\pm}\, $$10.7	53.0$$ \,{\pm}\, $$11	59.2$$ \,{\pm}\, $$12.2	<0.001b, c
Education, n (%)
Primary school	623 (11.8)	275 (13.1)	118 (8.8)	230 (12.7)	<0.001a-c
Secondary school	2323 (44.1)	1039 (49.4)	522 (39.0)	762 (41.9)
High school or above	2314 (44.0)	790 (37.5)	699 (52.2)	825 (45.4)
BMI, mean (SD), kg/m2	24.9$$ \,{\pm}\, $$3.3	25.0$$ \,{\pm}\, $$3.4	25.0$$ \,{\pm}\, $$3.2	24.8$$ \,{\pm}\, $$3.1	0.025b, c
Current smoker, n (%)	1674 (31.8)	819 (38.9)	365 (27.3)	490 (27.0)	<0.001a, b
Current drinker, n (%)	754 (14.3)	359 (17.1)	169 (12.6)	226 (12.4)	<0.001a, b
SBP, mean (SD), mmHg	131.0$$ \,{\pm}\, $$20.0	130.2$$ \,{\pm}\, $$19.1	129.1$$ \,{\pm}\, $$19.6	133.4$$ \,{\pm}\, $$20.9	<0.001b, c
DBP, mean (SD), mmHg	82.8$$ \,{\pm}\, $$11.1	82.3$$ \,{\pm}\, $$11.2	82.7$$ \,{\pm}\, $$10.9	82.5$$ \,{\pm}\, $$11.0	0.055
History of disease, n (%)
Hypertension	1326 (25.2)	499 (23.7)	312 (23.3)	515 (28.3)	<0.001b, c
Diabetes	400 (7.6)	137 (6.5)	83 (6.2)	180 (9.9)	<0.001b, c
Hyperlipidemia	578 (11.0)	192 (9.1)	162 (12.1)	224 (12.3)	0.001a, b

• Note: Categorical variables were represented as number (percentage); continuous variables were represented as mean ± standard deviation (SD).
• BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure.
• The p value is the results of overall comparisons, where p value < 0.05 indicates the significant difference of characteristics among the intensity of three groups, that is at least two groups are significantly different. The superscript indicates the detailed significant groups under significant level 0.05, where:
• ^a represents low versus moderate group.
• ^b represents low versus high group.
• ^c represents moderate versus high group.

Based on Kaplan–Meier survival analysis results, the KM-survival curve was significantly different among groups of physical activity intensity, as shown in Figure 2A. And the cumulative hazard in Figure 2B also significantly different between three intensity of physical activity groups. The p value for the overall and pairwise comparison were calculated by using log-rank test, and the results can be found in Figure 2C. According to the p values of pairwise comparison, we can find that Figures 2A and B also indicated that the moderate intensity of physical activity group had the longest median survival time, meanwhile it has the lowest hazard ratio compared with the other two groups at the same time. The cumulative hazard ratio of the high-intensity physical activity group is statistical significantly higher than other two groups, and its survival time is significantly less than the other two groups. The survival time and the cumulative hazard rates between the low and moderate groups were not significantly different. The results of baseline characteristics of participants by physical activity suggested importance of physical activity in the study of CVD events.

2.4 PA equation validation

In the validation cohort, we compared the PA equation with the China-PAR equation. As shown in Figures 3A and B, the AUCs for the PA equations were as good as China-PAR in both men and women, in particular, the AUC by PA equations was significantly higher than China-PAR in men (p value = 0.005) (Figure 3A). This result indicated that the PA equations were better fitted than the China-PAR in the physical activity cohort for Men. In addition, to further validate performance of PA equations, we divided participants into four categories (<5%, ≤5−<7.5%, ≤7.5−<10%, and ≥10%) based on the predicted risks. Categories were verified clinically meaningful cut points.¹⁴ Figure 4 were the comparison between Kaplan–Meier observed and predicted CVD event rates using PA equations. Histogram showed that predicted risk ratios were very close to observed risks in each cluster in both men and women. (Figures 4A and B). Overall, the results of AUCs and the histogram of predicted risks underscore the accuracy of our method, and the validation results also shown that compared with China-PAR model, our sex-specific PA equations were more applicable to the physical activity cohorts, especially in men.

3 DISCUSSION

This study investigated the physical activity cohort from Kailuan, China. Major findings of the present work are as follows. First, we found that physical activity was significantly associated with positive cardiovascular health outcome. Subjects who conducted moderate intensity of physical activity have the lower hazard ratio for CVD at the same time points comparing with the low and high group. On the contrary, the moderate group has longer survival time than the other two groups. Second, we developed physical activity equations (PA equations) for the Kailuan physical activity cohorts. From the PA equations, we found that physical activity reduced the hazards ratios of CVD.

The primary difference between our PA equations and prior published equations is, that we included physical activity into the risk estimating equation. Variables used in our model were age, SBP, CHOL, HDLC, current smoking state (yes/no), diabetes mellitus (yes/no), and physical activity. Of note, these variables are similar to those that have been used in other prediction models, except for the factor of physical activity. The values of the parameters in PA equations were close to China-PAR equations, which were developed based on the Chinese population. Our data show that PA equations perform better in physically active cohorts than China-PAR for Men.

In our study, physical activity was assessed using an international questionnaire short form, which is slightly different from the Harvard Alumni Questionnaire^{16, 28} and other physical activity questionnaires.²⁹ We classified subjects into three classes based on amount of physical activity times and frequencies per week. The amount of physical activity influenced risk prediction of CVD in this cohort. Previous studies have shown an association between the intensity of physical activity and the risk for the development of heart disease and CVD.^{30, 31} The American Heart Association and the American College of Cardiology give the recommendation to adults to engage into moderate-intensity physical activity around 150 min per week of or high-intensity physical activity around minutes per week.¹¹ The 2018 American Physical Activity Guidelines also provide an amount of physical activity for people of different age groups.³² In China, studies about the amount of physical activity have been scare, and in 2018, Shanghai Children's Medical Center released the first Physical activity guidelines for Chinese children and adolescents.³³ Hence, our study on the association between CVD risk and the amount of physical activity provides valuable guidance for Chinese patients regarding primary prevention of CVD.

Several limitations and deficiencies of the present study should be highlighted. First, in this work, we only performed internal cross validation without external validation of PA equation because of the limitation of cohort. Hence, we need to be careful when applying the PA equations to other populations. Therefore, in future studies, we will further validate PA equations in more physical activity cohorts. Second, the use of the Kailuan population limits this study; currently, the participants were only followed-up 9 years on average, and the size of subjects used in this study was not very big, particularly in a primary prevention trial with low outcome rates. The size may also have effect on the bias of the prediction model.⁷ Third, in this study, we used “IPQA” to measure physical activity levels, and it will be important to add objective criteria such as objectively measured physical activity by using accelerometer in the future. Last but not the least, during the 9-year follow-up, the intensity of physical activity may have changed for each participant and thus, physical activity can be a dynamic variable that changes over time. In our model, we did not account for changes in the physical activity and the subsequent risk of CVD. Besides, it should be stressed that the PA equation is better in men than women. From Table 3, it is obvious that most of the factors’ coefficients of PA equation were significant in men. In contrast, several factors’ coefficients were not significant in women. There could be many reasons for this, such as the gender difference in CVD.³⁴ It has been documented that age, CHOL and HDLC have strong effects in men, but diabetes and smoking have main influence in women. However, the baseline characteristics in Table 1 showed that the percentage of women who smoke is very low in our cohort. Hence, the equation for predicting risk of CVD might not be as robust or reliable in female population. In addition, the intensity of physical activity was significantly different between men and women. The number of women with different intensity of physical activity was relatively balanced (as shown in Table 2). By contrast, the proportion of men with different intensity of physical activity was various (as shown in Table 2). Nevertheless, we developed PA equations for physical activity cohorts, which provides an insight on the development of prediction models for CVD, which will have clinical values in the future.

In conclusion, we developed a prediction model for a physically active cohort based on a Cox proportional hazard regression. Our developed sex-specific PA equations have shown good performance to predict CVD for a physically active cohort in the Kailuan population. The results suggest that physical activity has an inverse association with CVD risk in this population.

4 METHODS

AUTHOR CONTRIBUTION

J. X. and Y. Z. conceived the protocol. J. Z., Y. Y., X. Z., Y. X., S. A., X. D., and F. H. contributed to the analysis and interpretation of data. J. Z. drafted the manuscript. J. X., Y. Z., H. L., and G. L. critically revised the manuscript. All authors agree to be fully accountable for ensuring the integrity and accuracy of the work and read and approved the final manuscript. The corresponding author had full access to all data in the study and assumed final responsibility for the decision to submit the manuscript for publication.

CONFLICT OF INTEREST STATEMENT

The authors declare no competing interest.

ETHICS STATEMENT

The study was conducted on the basis of the guideline of the Declaration of Helsinki and was approved by the Ethics Committees of the Kailuan General Hospital, Beijing Tongren Hospital, and Beijing Tiantan Hospital. The study was also approved by the Ethics Committee of Shanghai University (2021-195). Written informed consent was taken from all participants at the time of study inclusion.